MCP Toolkit

MCP Toolkit

A unified MCP server that provides tools for UI/UX design intelligence, command and agent libraries, and development skills, enabling enhanced design, development, and workflow assistance.

Category
访问服务器

README

MCP Toolkit

Overview

A unified Python MCP (Model Context Protocol) server built on FastMCP that exposes three integrated toolkits to Claude Desktop, Claude Code, and claude.ai: UI/UX Pro Max design intelligence, Claude Command Suite (slash commands + AI agents), and Superpowers development skills. One server, one config, 14 tools spanning design-system generation, stack-specific UI guidelines, browsable command/agent libraries, and composable development skills (TDD, debugging, code review, planning).

Status

Working — server runs via stdio, SSE, or streamable-HTTP; start.sh publishes it through a Cloudflare tunnel for claude.ai integration.

14 tools exposed:

Group Tools
UI/UX Pro Max uiux_search, uiux_design_system, uiux_stack_guidelines, uiux_list_domains
Command Suite list_commands, get_command, search_commands, list_agents, get_agent
Superpowers list_skills, get_skill, get_skill_file, search_skills, get_workflow

Content inventory:

  • Commands: 26 namespaces (dev, test, deploy, security, project, rust, svelte, simulation, docs, team, performance, memory, reasoning, wfgy, etc.)
  • Agents: 40+ specialized agents (code-auditor, test-engineer, release-manager, svelte-*, swift-macos-expert, wfgy, skill-builder, data-ai, infrastructure, …)
  • Skills: 31 Superpowers SKILL.md folders (brainstorming, systematic-debugging, test-driven-development, writing-plans, frontend-design, mcp-builder, webapp-testing, …)
  • UI/UX data: CSVs for styles, 161 color palettes, typography pairings, charts, product types, UX guidelines, React/web patterns, Google Fonts, prompt templates — plus react-native stack guidelines

Setup & Run

# Install (only FastMCP required)
pip install -r requirements.txt

Claude Desktop (stdio)

fastmcp run server.py

Add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "mcp-toolkit": {
      "command": "python3",
      "args": ["-m", "fastmcp", "run", "/absolute/path/to/mcp-toolkit/server.py"]
    }
  }
}

Claude Code

claude mcp add mcp-toolkit -- python3 -m fastmcp run /absolute/path/to/server.py

claude.ai (remote via Cloudflare tunnel)

./start.sh              # server on :8100 + cloudflared tunnel (prints trycloudflare.com URL)
./start.sh --no-tunnel  # local-only

Then in claude.ai: Settings → Integrations → Add Custom MCP → paste the https://…trycloudflare.com/mcp URL.

Manual alternative: fastmcp run server.py -t sse --port 8100 or -t streamable-http.

Architecture

mcp-toolkit/
├── server.py          # FastMCP server, 14 @mcp.tool() definitions (UIUX / Commands / Skills)
├── start.sh           # Boots streamable-http server + Cloudflare tunnel, prints public URL
├── requirements.txt   # fastmcp>=3.0.0
├── agents/            # 40+ agent .md files (flat + subdirs: development, security, business, …)
├── commands/          # 26 namespaces of slash-command .md files (dev, test, deploy, rust, svelte, …)
├── skills/            # 31 Superpowers skills, each a folder with SKILL.md + supporting files
├── uiux_data/         # CSVs: styles, colors, typography, charts, products, UX, stacks
└── uiux_scripts/      # core.py, design_system.py, search.py — imported by server.py

Wiring note: server.py patches sys.path and core.DATA_DIR so the bundled uiux_scripts/ modules resolve uiux_data/ regardless of cwd.

Tool surface:

  • uiux_search(query, domain, max_results) — 12 domains auto-detected
  • uiux_design_system(query, project_name) — full design-system markdown (colors/typography/spacing/components)
  • uiux_stack_guidelines(query, stack) — currently react-native
  • list_commands / get_command / search_commands — walk commands/<ns>/*.md
  • list_agents / get_agent — walk agents/**/*.md, parse frontmatter description:
  • list_skills / get_skill / get_skill_file / search_skills — walk skills/*/SKILL.md
  • get_workflow(workflow) — curated multi-skill recipes for feature | bugfix | refactor | review

Roadmap

  • [ ] Expand uiux_stack_guidelines beyond react-native (React, Next.js, Vue, Svelte, SwiftUI, Flutter, Tailwind, shadcn/ui, HTML/CSS)
  • [ ] Cache _get_agents() / _get_namespaces() / SUPERPOWERS walks — currently re-scan disk on every tool call
  • [ ] Make SUPERPOWERS set construction lazy; it runs at import time and will crash if skills/ is ever missing
  • [ ] Replace hardcoded Python path in start.sh (/Library/Frameworks/Python.framework/Versions/3.11/bin/python3) with a venv or $(command -v python3)
  • [ ] Auth / access control for the public Cloudflare tunnel endpoint
  • [ ] Health check + reconnect logic in start.sh (currently fixed sleep 8 for tunnel URL parse)
  • [ ] Add MCP resources (not just tools) for large static content — design-system CSVs, agent/skill catalogs
  • [ ] Tests — no test suite exists yet

Known Bugs

  • None identified in shipped tool code. (TODO markers found in commands/skills/scripts/init_skill.py and quick_validate.py are intentional template scaffolding for the skill-builder, not defects.)
  • Minor: SUPERPOWERS = {…} at module top-level will raise FileNotFoundError if skills/ is absent — deployment-time concern only.
  • Minor: start.sh parses the tunnel URL after a fixed 8-second sleep; slow cloudflared startups may leave TUNNEL_URL empty.

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选